TSMC has ushered in a new two-nanometer era.

      The leading chipmaker, TSMC, has begun mass production of chips using a 2‑nanometer process. Simply because that’s what was written in the plan. Without fanfare, but with unfailing punctuality, the production lines at Fab 22 in Kaohsiung began stamping out the future.

      In a world that expects breakthroughs every year, TSMC’s engineers offer a modest but charming piece of arithmetic. Their new N2 technology is either 10–15% more speed at the same appetite for power, or 25–30% less heat and power consumption while maintaining performance. Transistor density has increased by roughly a fifth. That doesn’t sound as epic as a “revolution,” but in a world where Moore’s Law has long been limping, every such improvement is a small miracle. The secret is the shift to a new type of transistor — GAA (Gate‑All‑Around), which, as the name implies, “hugs” the channel from all sides for better control. TSMC has neatly borrowed this idea from its rival Samsung.

      Who will get these chips?

      While TSMC was ramping up production, South Korea’s Samsung also announced its 2 nm in December. The difference, as insiders like to say, is in the “yield” — the percentage of perfect chips from a wafer. Rumor has it TSMC’s was immediately closer to 80%, whereas Samsung, according to some reports, is only approaching 60% with its second‑generation process. This is that boring manufacturing magic that decides everything.

      Meanwhile, the main question is who will get these precious wafers. The answer, as almost always, is Apple. The folks in Cupertino will traditionally be the first and most important customers. But this time everyone building AI systems is also scrambling for a piece of the pie, and demand for the newcomer is so large that TSMC’s capacity for 2026 is already completely sold out. The irony is that while the most advanced chips will be born only in Taiwan, TSMC is busily building fabs in Arizona, USA. However, by local standards they will be a bit “outdated” — 3 nm, and only by 2027. That leaves a window of opportunity for Samsung, which can offer 2 nm to those American customers who don’t want to wait or worry about logistics.

      Quiet work on the next product

      TSMC isn’t about to rest on its laurels. In 2026 the company plans to release an improved N2P, and is also preparing a revolution codenamed A16 — a technology that brings power to the chip from the backside. This should provide another significant leap in efficiency. It seems the only things produced louder and faster than the chips themselves are roadmaps and the ambitions of semiconductor giants. All that remains for us is to wait for these quiet successes to turn into the next, slightly smarter and slightly less power‑hungry gadget.

      Meanwhile, shares of Chinese AI‑chip maker Moore Threads soared more than sixfold on the day of its IPO on the Shanghai Stock Exchange. Behind the phenomenal frenzy lies not just speculative fever, but a belief in technological sovereignty and in the success of a company often called the “Chinese Nvidia.”